The rs429358 locus in apolipoprotein E is associated with hepatocellular carcinoma inpatients with cirrhosisInnes, Hamish; Nischalke, Hans Dieter; Guha, Indra Neil; Weiss, Karl Heinz; Irving, Will;Gotthardt, Daniel ; Barnes, Eleanor; Fischer, Janett; Ansari, M. Azim; Rosendahl, Jonas; Lin,Shang-Kuan; Marot, Astrid; Pedergnan, Vincent; Casper, Markus; Benselin, Jennifer;Lammert, Frank; McLauchlan, John; Lutz, Philip L.; Hamill, Victoria; Mueller, Sebastian;Morling, Joanne R.; Semmler, Georg; Eyer, Florian; Felden, Johann von; Link, Alexander;Vogel, Ardnt; Marquardt, Jens U.; Sulk, Stefan; Trebicka, Jonel; Valenti, Luca; Datz,Christian; Reiberger, Thomas; Schafmayer, Clemens; Berg, Thomas; Deltenre, Pierre;Hampe, Jochen; Stickel, Felix; Buch, StephanPublished in:Hepatology CommunicationsDOI:10.1002/hep4.1886Publication date:2021Document VersionPublisher's PDF, also known as Version of recordLink to publication in ResearchOnlineCitation for published version (Harvard):Innes, H, Nischalke, HD, Guha, IN, Weiss, KH, Irving, W, Gotthardt, D, Barnes, E, Fischer, J, Ansari, MA,Rosendahl, J, Lin, S-K, Marot, A, Pedergnan, V, Casper, M, Benselin, J, Lammert, F, McLauchlan, J, Lutz, PL,Hamill, V, Mueller, S, Morling, JR, Semmler, G, Eyer, F, Felden, JV, Link, A, Vogel, A, Marquardt, JU, Sulk, S,Trebicka, J, Valenti, L, Datz, C, Reiberger, T, Schafmayer, C, Berg, T, Deltenre, P, Hampe, J, Stickel, F & Buch,S 2021, 'The rs429358 locus in apolipoprotein E is associated with hepatocellular carcinoma in patients withcirrhosis', Hepatology Communications. rightsCopyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright ownersand it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.Take down policyIf you believe that this document breaches copyright please view our takedown policy at for detailsof how to contact us.Download date: 22. Feb. 2022

Hepatology Communications, VOL. 0, NO. 0, 2021 The rs429358 Locus in ApolipoproteinE Is Associated With HepatocellularCarcinoma in Patients With CirrhosisHamish Innes,1-3 Hans Dieter Nischalke,4 Indra Neil Guha,5 Karl Heinz Weiss,6 Will Irving,5 Daniel Gotthardt,7Eleanor Barnes,8 Janett Fischer,9 M. Azim Ansari,8 Jonas Rosendahl,10 Shang- Kuan Lin,8 Astrid Marot,11,12 Vincent Pedergnana,13Markus Casper,14 Jennifer Benselin,5 Frank Lammert,14 John McLauchlan,15 Philip L. LutzSebastian Mueller,16,17 Joanne R. Morling,2,5 Georg SemmlerAlexander Link,22 Arndt Vogel,23 Jens U. MarquardtChristian Datz,19 Thomas Reiberger25,33Jochen Hampe,34,18,19 Florian Eyer,4 Victoria Hamill,1,3,20 Johann von Felden,24 Stefan Sulk,25 Jonel Trebicka,26,27 Luca Valenti,18 Clemens Schafmayer,30 Thomas Berg,9 Pierre Deltenre25* Felix Stickel, * and Stephan Buch *,21,28,29,11,31,32*The host genetic background for hepatocellular carcinoma (HCC) is incompletely understood. We aimed to determineif four germline genetic polymorphisms, rs429358 in apolipoprotein E (APOE), rs2642438 in mitochondrial amidoximereducing component 1 (MARC1), rs2792751 in glycerol- 3- phosphate acyltransferase (GPAM), and rs187429064 in transmembrane 6 superfamily member 2 (TM6SF2), previously associated with progressive alcohol- related and nonalcoholicfatty liver disease, are also associated with HCC. Four HCC case- control data sets were constructed, including two mixedetiology data sets (UK Biobank and FinnGen); one hepatitis C virus (HCV) cohort (STOP- HCV), and one alcohol- related HCC cohort (Dresden HCC). The frequency of each variant was compared between HCC cases and cirrhosiscontrols (i.e., patients with cirrhosis without HCC). Population controls were also considered. Odds ratios (ORs) associations were calculated using logistic regression, adjusting for age, sex, and principal components of genetic ancestry. Fixed- effect meta- analysis was used to determine the pooled effect size across all data sets. Across four case- control data sets,2,070 HCC cases, 4,121 cirrhosis controls, and 525,779 population controls were included. The rs429358:C allele (APOE)was significantly less frequent in HCC cases versus cirrhosis controls (OR, 0.71; 95% confidence interval [CI], 0.61- 0.84;P 2.9 10 5). Rs187429064:G (TM6SF2) was significantly more common in HCC cases versus cirrhosis controls andexhibited the strongest effect size (OR, 2.03; 95% CI, 1.45- 2.86; P 3.1 10 6). In contrast, rs2792751:T (GPAM) wasnot associated with HCC (OR, 1.01; 95% CI, 0.90- 1.13; P 0.89), whereas rs2642438:A (MARC1) narrowly missedstatistical significance (OR, 0.91; 95% CI, 0.84- 1.00; P 0.043). Conclusion: This study associates carriage of rs429358:C(APOE) with a reduced risk of HCC in patients with cirrhosis. Conversely, carriage of rs187429064:G in TM6SF2 is associated with an increased risk of HCC in patients with cirrhosis. (Hepatology Communications 2021;0:1-14).Hepatocellular carcinoma (HCC) is the thirdmost common type of cancer death, responsible for approximately 800,000 deathsglobally every year worldwide.(1) Most cases of HCCdevelop against a background of advanced liver fibrosis and cirrhosis. Like any cancer, HCC is a productAbbreviations: HU, Hounsf ield unit; APOA/B/E, apolipoprotein A/B/E; ArLD, alcohol- related liver disease; AST, aspartate aminotransferase;BMI, body mass index; CI, conf idence interval; CVD, cardiovascular disease; FE, f inite element; GPAM, glycerol- 3- phosphate acyltransferase; HCC,hepatocellular carcinoma; HCV, hepatitis C virus; HSD17B13, 17- β hydroxysteroid dehydrogenase 13; ICD- 9/10, International Classif icationof Diseases, Ninth/Tenth Revision; LDLR, lipoprotein receptor; LOR, log odds ratio; MARC1, mitochondrial amidoxime reducing component 1;NAFLD, nonalcoholic fatty liver disease; OR, odds ratio; TM6SF2, transmembrane 6 superfamily member 2; UKB, United Kingdom Biobank.Received September 1, 2021; accepted November 24, 2021.Additional Supporting Information may be found at info.*These authors contributed equally to this work.Supported by the Medical Research Foundation (Viral Hepatitis Fellowship grant C0825 to H.I.), Swiss National Funds (No. 310030 169196to F.S.), Swiss Foundation for Alcohol Research (No. 261/15 to F.S.), German Federal Ministry for Education and Research, Liver SystemsMedicine Network (No. 031L0031 to J.H.), Medical Research Council (No. MR/K01532X/1 to B., Clinician Scientist Fellowship grant MR/P008348/1 to J.R.M, No. MC UU 12014/1 to J.M.), Medical Research Foundation (No. C0365 to J.M.), Deutsche Krebshilfe (No. 701121691

INNES ET AL.of somatic mutations acquired in pivotal driver genesbut also influenced by germline (i.e., constitutional)polymorphisms modifying the susceptibility toHepatology Communications, Month 2021developing HCC.(2) Thus far, several such polymorphisms have been identified and robustly validated,including rs738409 in patatin- like phospholipaseto H.D.N.), European Commission European Funds for Regional Development and Regional Ministry of Economy, Science, and Digitalization (No.ZS/2018/11/95324 to A.L.), Deutsche Forschungsgemeinschaft (SFB TRR57 to P18, CRC 1382 A09 to J.T.), European Union’s Horizon 2020Research and Innovation Programme (Galaxy, No. 668031 to J.T.; MICROB- PREDICT, No. 825694 to J.T.; DECISION, No. 84794 to J.T.; No.731875 to J.T.; No. 777377 to L.V.; Photonics, No. 101016726 to L.V.; and Gilead IN- IT- 989- 5790 to L.V.), Cellex Foundation (PREDICT toJ.T.), MyFirst Grant AIRC (No. 16888 to L.V.), Ministero della Salute (No. RF- 2016- 02364358 to L.V.), Fondazione IRCCS (No. PR- 0391 toL.V., No. RC100017A to L.V.), and Cancer Research UK (No. C30358/A29725 to E.B.).The views expressed in this article are those of the authors and not necessarily those of the National Health Service, the NIHR, or the Department ofHealth. The funders had no involvement in the study design; collection, analysis and interpretation of data; writing of the report; and decision to submitthe article for publication. 2021 The Authors. Hepatology Communications published by Wiley Periodicals LLC on behalf of American Association for the Study of LiverDiseases. This is an open access article under the terms of the Creat ive Commo ns Attri bution License, which permits use, distribution and reproductionin any medium, provided the original work is properly cited.View this article online at 10.1002/hep4.1886Potential conflict of interest: Dr. Trebicka has received speaking and/or consulting fees from Gore, Bayer, Alexion, MSD, Gilead, Intercept,Norgine, Grifols, Versantis, and Martin Pharmaceutical. Dr. Valenti has received speaking fees from MSD, Gilead, AlfaSigma, and AbbVie; hehas served as a consultant for Gilead, Pf izer, AstraZeneca, Novo Nordisk, Intercept, Diatech Pharmacogenetics, and Ionis Pharmaceuticals andreceived research grants from Gilead. The other authors have nothing to report.ARTICLE INFORMATION:From the 1School of Health and Life Sciences, Glasgow Caledonian University, Glasgow, United Kingdom; 2Population and LifespanSciences, School of Medicine, University of Nottingham, Nottingham, United Kingdom; 3Public Health Scotland, Glasgow, United Kingdom;4Department of Internal Medicine I, University Hospital, University of Bonn, Bonn, Germany; 5National Institute for Health Research(NIHR), Nottingham Biomedical Research Centre, Nottingham University Hospitals National Health Service Trust and the University ofNottingham, Nottingham, United Kingdom; 6Department of Gastroenterology and Hepatology, University Hospital Heidelberg, Heidelberg,Germany; 7Department of Internal Medicine IV, Medical University of Heidelberg, Heidelberg, Germany; 8Peter Medawar Building forPathogen Research, Nuffield Department of Medicine and the Oxford NIHR Biomedical Research Centre, Oxford University, Oxford,United Kingdom; 9Division of Hepatology, Department of Medicine II, Laboratory for Clinical and Experimental Hepatology, LeipzigUniversity Medical Center, Leipzig, Germany; 10Medical Department 1, University Hospital Halle, Martin- Luther Universität Halle- Wittenberg, Halle, Germany; 11Division of Gastroenterology and Hepatology, Centre Hospitalier Universitaire Vaudois, Université deLausanne, Lausanne, Switzerland; 12Department of Gastroenterology and Hepatology, Centre Hospitalier Universitaire, UCLouvainNamur, Université Catholique de Louvain, Yvoir, Belgium; 13Laboratoire MIVEGEC, Montpellier, France; 14Department of MedicineII, Saarland University Medical Center, Saarland University, Homburg, Germany; 15Medical Research Council- University of GlasgowCentre for Virus Research, Glasgow, United Kingdom; 16Center for Alcohol Research, University of Heidelberg, Heidelberg, Germany;17Medical Department, Salem Medical Center, Heidelberg, Germany; 18Department of Internal Medicine III, Division of Gastroenterologyand Hepatology, Medical University of Vienna, Vienna, Austria; 19Department of Internal Medicine, General Hospital Oberndorf, TeachingHospital of the Paracelsus Medical University Salzburg, Oberndorf, Austria; 20Department of Clinical Toxicology, Klinikum Rechts derIsar, Technical University of Munich, Munich, Germany; 21Department of Medicine, University Medical Center Hamburg- Eppendorf,Hamburg, Germany; 22Department of Gastroenterology, Hepatology, and Infectious Diseases, Otto- von- Guericke University Hospital,Magdeburg, Germany; 23Department of Gastroenterology, Hepatology, and Endocrinology, Hannover Medical School, Hannover, Germany;24Department of Medicine I, University Hospital Schleswig Holstein– Campus Lübeck, Lübeck, Germany; 25Medical Department 1, UniversityHospital Dresden, Technische Universität Dresden, Dresden, Germany; 26Department of Internal Medicine I, Goethe University, Frankfurt,Germany; 27European Foundation for Study of Chronic Liver Failure, Barcelona, Spain; 28Precision Medicine– Department of TransfusionMedicine and Hematology, Fondazione IRCCS Ca’ Granda Ospedale Maggiore Policlinico, Milan, Italy; 29Department of Pathophysiologyand Transplantation, Università degli Studi di Milano, Milan, Italy; 30Department of General, Visceral, Vascular, and TransplantSurgery, Rostock University Medical Center, Rostock, Germany; 31Department of Gastroenterology, Hepatopancreatology, and DigestiveOncology, University Clinics of Brussels Hospital Erasme, Brussels, Belgium; 32Department of Gastroenterology and Hepatology, CliniqueSt Luc, Bouge, Belgium; 33Center for Regenerative Therapies Dresden, Technische Universität Dresden, Dresden, Germany; 34Departmentof Gastroenterology and Hepatology, University Hospital of Zurich, Zurich, Switzerland.ADDRESS CORRESPONDENCE AND REPRINT REQUESTS TO:Hamish Innes, Ph.D.School of Health and Life SciencesGlasgow Caledonian UniversityCowcaddens Road2G40BA, Glasgow, United KingdomE-mail: [email protected]: 44 141 331 3000

Hepatology Communications, Vol. 0, No. 0, 2021domain containing 3 (PNPLA3), rs58542926 intransmembrane 6 superfamily member 2 (TM6SF2),and rs72613567 in 17- β hydroxysteroid dehydrogenase 13 (HSD17B13).(3- 6) However, this explains onlypart of the host genetic background underlying HCCdevelopment. A more complete understanding of theconstitutional genetic polymorphisms that predispose patients to HCC could herald several important advancements. In the short term for example, itcould support risk stratification of patients with cirrhosis with respect to HCC screening decision(7); inthe longer term, it could guide the discovery of chemoprevention agents if any of the risk loci prove to be“druggable.”In a recent exome association study, Jamialahmadiet al.(8) identified three novel missense variants associated with hepatic fat content. These variants werers429358 in apolipoprotein E (APOE, rs2792751 inglycerol- 3- phosphate acyltransferase (GPAM), andrs187429064 in TM6SF2, where the latter is incomplete linkage equilibrium with the better knownrs58542926 locus. They also show that the rs2642438missense variant in mitochondrial amidoxime reducing component 1 (MARC1), which we and others haverecently identified as a risk factor for cirrhosis,(9,10)is associated with liver fat content, too. In a parallelstudy, Bianco et al.(11) indicated that higher liver fatcontent is causally associated with HCC occurrence.On that basis, genetic factors that alter liver fat content may also alter HCC risk; these variants therefore warrant exploration in candidate gene- associationstudies for HCC. To that end, our primary objectivewas to explore a possible association of each of thesefour aforementioned variants with HCC across a variety of large data sets and etiologies.Materials and MethodsSCIENTIFIC APPROACHThis study uses data from the following four HCCcase- control data sets: two mixed etiology cohorts(United Kingdom Biobank [UKB] and FinnGen),one hepatitis C virus (HCV) cohort (STOP- HCV),and one alcohol- related liver disease (ArLD) cohort(Dresden study).The following four candidate variants were considered for association with HCC: 1) rs429358 (APOE), 2)INNES ET AL.rs2792751 (GPAM), 3) rs2642438 (MARC1), and 4)rs187429064 (TM6SF2). Genotyping methods forthese variants are described in Supporting MaterialsAppendix A.In the broadest terms, our goal was to assess if thefrequency of these variants was different for HCCcases versus non- HCC controls. Two types of non- HCC controls were considered, patients with cirrhosiswithout HCC and population controls without HCC.On one hand, comparing HCC cases to cirrhosis controls is essential to eliminate confounding, i.e., becausevariants associated with HCC tend also to be associated with progression to cirrhosis. On the other hand,population controls lend insight because HCC canalso arise in patients at a precirrhosis stage (i.e., particularly nonalcoholic fatty liver disease [NAFLD]- related HCC).(12,13) A population perspective is alsorelevant to early detection case- finding initiatives forHCC.(14)CASE- CONTROL DATA SETSUK BiobankThe UKB is a cohort of half a million middle- agedindividuals from the United Kingdom, recruited in2006- 2010. Blood specimens donated at enrollmenthave been used to characterize participants in termsof genetic factors as well as being serum biomarkers(e.g., alanine aminotransferase). Participant data arealso linked to UK health registries to capture medical presentations occurring both before and afterenrollment.(15)This study was restricted to UKB participantsof White British ancestry (UKB field ID: 22006).We then excluded those with a poor quality geneticsample (defined by UKB field ID: 22027) or whowere related to another participant (inferred bya kinship coefficient 0.1). Cases were participants with a history of HCC, defined as a hospitaladmission, death, or cancer registration with HCC(International Classification of Diseases, TenthRevision [ICD- 10]: C22.0, or ICD- 9: 155.0), eitherbefore or after UKB enrollment. Liver disease etiology for the HCC cases was estimated using ahierarchical definition of a) viral hepatitis, b) autoimmune liver disease in the absence of a, c) ArLD inthe absence of a- b, d) NAFLD in the absence of a- c,and e) other/unknown in the absence of a- d. Risk3

INNES ET AL.factors for these etiologies were discerned through acombination of hospital admissions and/or information reported during the UKB enrollment interview(Supporting Table S1).The following two control groups were considered:ontrol group 1 included UKB participants with a hospital admission for liver cirrhosis but without a historyof HCC. Hospital admissions due to cirrhosis wereidentified using a validated set of ICD and operation/procedure codes(16) (See Supporting Table S2 for further details.) Control group 2 included all UKB participants without a history of HCC. The vast majorityof individuals in this group had no history of chronicliver disease. Control group 2 was broadly equivalentto a general population control group.FinnGenFinnGen is a public– private partnership project, combining genotyping data from Finnish biobanks with electronic health record data derived fromnational health registries. Genome- wide associationstudy (GWAS) summary statistics for more than1,800 phenotypes/endpoints, including for primaryliver cancer, have been publically released.For this study, we used the latest R4 data released(published November 2020) pertaining to a samplesize of 176,899 individuals.(17) Cases were individuals with a history/diagnosis of primary liver cancer(ICD- 10: C22 and ICD- 9: 155), whereas controlswere all individuals without a diagnosis of primaryliver cancer. Similar to the UKB control group 2, thislargely comprised individuals without any preexistingliver disease. GWAS summary statistics relating specifically to HCC were not available.STOP- HCV CIRRHOSIS STUDYThe STOP- HCV cirrhosis study comprisedapproximately 1,200 patients with hepatitis C- relatedcirrhosis. Participants were recruited from 31 specialistliver clinics in the United Kingdom between January2015 and July 2016. Cirrhosis was defined throughhistologic assessment, imaging, or a validated serumbiomarker consistent with liver cirrhosis (i.e., aspartate aminotransferase [AST]- to- platelet ratio index 2, FibroTest 0.73, or enhanced liver fibrosis score 10.48). Blood specimens collected at enrollmentwere used to generate host- genotyping data through4Hepatology Communications, Month 2021the Affymetrix UK Biobank array. Furthermore, participants from England have been linked to nationalhospital admission, cancer registrations, and mortalitydata.The present analysis was restricted to participantsfrom England (i.e., to ensure complete data on hospital admissions, cancer registrations, and mortality)and participants of White ethnicity. As with the UKB,cases were defined on the basis of an in- patient hospital admission, death, or cancer registration indicatingHCC (ICD- 10: C22.0; ICD- 9: 155.0) before or afterstudy enrollment. Controls were all participants without a history of HCC.DRESDEN ALCOHOL HCCCOHORTThe Dresden HCC cohort included 2,311 patientswith a history of high- risk alcohol consumption inwhom nonalcohol- related causes of chronic liver disease had been excluded. Patients were recruited fromgastroenterology and hepatology hospitals across fiveEuropean countries (Austria, France, Germany, Italy,and Switzerland). For this study, cases were patientswith a diagnosis of HCC determined through histologic and/or imaging (computed tomography or magnetic resonance imaging [MRI]) investigations.As with the UKB, two control groups were considered. Control group 1 was individuals diagnosedwith alcohol- related cirrhosis but without a historyof HCC. Control group 2 comprised patients withoutcirrhosis.The diagnosis of alcohol- related cirrhosis wasestablished as described in detail.(18) Briefly, thediagnosis was based on a history of prolonged sustained alcohol intake of a minimum of 40 g/day inwomen and 60 g/day in men, together with histologicexamination of liver tissue or compatible historical,clinical, laboratory, radiologic, and endoscopic features of advanced chronic liver disease. Patients wereexcluded if they had any other potential cause ofliver injury, specifically if they were positive for hepatitis B surface antigen, anti- HCV, antinuclear antibodies (titer 1:80), or antimitochondrial antibodies(titer 1:40). Patients with elevated serum ferritinconcentrations and a transferrin saturation 50%,a serum ceruloplasmin concentration 20 mg/dL(0.2 g/dL), or a serum alpha- 1 antitrypsin concentration 70 mg/dL (13 µmol/L) were further

Hepatology Communications, Vol. 0, No. 0, 2021investigated and excluded, as appropriate. The diagnosis of HCC was based on histologic examinationof tumor tissue or evidence on imaging, preferablyusing two modalities, of lesions that were hypervascular in the arterial phase with washout in the portalvenous or delayed phases.(19)Patients with alcohol misuse but no evidence of cirrhosis (control group 2) were recruited as described indetail.(15) In brief, these patients had a background ofalcohol consumption of at least 60 g/day for 10 yearswith or without features of alcohol dependence(20);none had historical, clinical, or laboratory evidence ofcirrhosis as reflected by AST- adapted cut- off valuesfor liver stiffness measured by transient elastography,as described.(21)All participants from the Dresden cohort were ofCaucasian ancestry, and genotyping was performedusing the Illumina BeadChip array (see SupportingMaterials Appendix A). The study protocol wasapproved by the ethics committees of the participating institutions, and all patients provided writteninformed consent before study inclusion.DATA ANALYSISAssociation With Liver Fat ContentWe started by replicating the UKB associationbetween each candidate variant and liver fat fraction, as reported by Jamialahmadi et al.(8) Thisallowed us to compare each variant’s direction ofassociation with liver fat content with the direction of association for HCC. Liver fat fractionwas measured through MRI, which at the time ofanalysis was available for a subset of 9,893 participants (UKB Field ID: 22436). We performed log10transformation on this variable to achieve approximate normality. Covariate adjustment was includedfor body mass index (BMI), age, sex, and the topfive principal components of genetic ancestry. Theanalysis was restricted to participants in the WhiteBritish ancestry subset (UKB Field ID: 22006).Association With HCCFor each candidate variant, we computed the simple minor allele frequencies (MAFs) in cases and controls from all four cohorts. The association betweeneach candidate variant and HCC was then quantifiedINNES ET AL.through multivariate logistic regression. All associations were adjusted for age, sex, and the top principal components of genetic ancestry. However, therewere minor differences by data set, which are outlined in Supporting Table S3. We did not controlfor established HCC risk variants (i.e., rs738409 inPNPLA3, rs58542926 in TM6SF2, and rs72613567in HSD17B13) because these were all in linkage equilibrium with the candidate variants considered (i.e.,R2 0.001). The exception to this was rs187429064where we included adjustment for the rs58542926genotype out of prudence, given that both variants liein TM6SF2. All associations were calculated underan additive genetic model, with two- tailed P valuespresented.We then performed a fixed- effect meta- analysisto determine a pooled effect size across studies,using the METAL software package.(22) Two pooledeffect sizes were calculated. First, a pooled effectsize specific to cirrhosis controls (i.e., UKB controls 1 Dresden controls 1 STOP- HCV ). Second,an overall effect size specific to population controls(i.e., UKB controls 1 FinnGen). All meta- analyseswere weighted according to the effective samplesize, defined according to the formula 4/(1/numberof cases 1/number of controls). A Bonferroni- corrected P 0.0125 was used to judge statisticalsignificance.Associations were expressed either in terms of logodds ratio (LOR) or odds ratios (ORs), where thelatter is simply the exponent of the LOR. In graphical figures, we present associations in terms of theirLORs because these are symmetrical around the nulland thus allow one to visually compare magnitude ofassociations for variants that affect risk in opposingdirections. Various polygenic risk scores were alsocreated, and their association with HCC was quantified (see Supporting Materials Appendix B).ResultsASSOCIATION WITH LIVER FATCONTENTAll four variants were strongly associated with liverfat content, with P values ranging from 2.1 10 6 to3.7 10 10 (see Fig. 1). Two of the four variants wereassociated with reduced liver fat content (rs429358:C5

INNES ET AL.Hepatology Communications, Month 2021FIG. 1. Association of candidate variants with liver fat content in the UKB APOE and rs2642438:A in MARC1), whereastwo variants were associated with increased liver fatcontent (rs2792751:T in GPAM and rs187429064:Gin TM6SF2). The rs187429064:G variant exhibited the strongest effect size (beta, 0.29), followedby rs429358:C (beta, 0.09), then rs279275:T (beta,0.06), and then rs2642438:A (beta, 0.05).CASE- CONTROL DATAIn total, the four case- control data sets included2,070 HCC cases and 4,121 cirrhosis controls. Overhalf the cases were alcohol- related HCCs from theDresden study (n 1,289), and 149 were hepatitisC- related HCCs from the STOP- HCV study. Therewere 366 HCC cases identified from the UKB study.Of these, we estimate that 153 (43%) were related toNAFLD, 115 (31%) related to ArLD, and 29 (9%)related to viral hepatitis. Cases were largely men(73%- 91%), with a mean age ranging from 60 to 69years depending on the study (Table 1).Of the cirrhosis controls, 691 were from STOP- HCV, 2,536 from the UKB, and 894 from Dresden.Cirrhosis controls were again predominantly men(63%- 77%) but were younger than HCC cases (meanage ranging from 55 to 59 years). The UKB andFinnGen non- liver disease control groups comprised349,018 and 176,633 FinnGen individuals, respectively (Table 1). Of the 128 noncirrhotic controls fromthe Dresden cohort, 50%, 40%, and 10% were estimated to be at Metavir stage F0, F1- 2, and F3, respectively, based on AST- adapted liver stiffness cutoffs.In the STOP- HCV cohort, about one fifth (19.6%)had achieved a hepatitis C sustained viral response atthe time of study enrollment. This proportion wascomparable for HCC cases (20.8%) and cirrhosis controls (19.4%).6ASSOCIATION WITH HCCAPOE (rs429358)The APOE rs429358:C allele was consistently lessfrequent in cases versus controls across all data sets.For example, 10.8% in UKB cases versus 13.9% innon- HCC controls (Table 1). In multivariate regression, rs429358:C was independently associated witha reduced HCC risk across all comparisons and datasets. The pooled OR for each copy of the rs429358:Callele was 0.71 (95% confidence interval [CI], 0.61- 0.84; P 2.9 10 5) against cirrhosis controls and0.66 (95% CI, 0.57- 0.78; P 1.0 10 6) against population controls (see Fig. 2).GPAM (rs2792751)The GPAM rs2792751:T allele was generallyhigher in HCC cases versus controls. However,the differences were modest; for example, 33.6% inDresden HCC cases versus 32.0% in cirrhosis controls. In multivariate regression, the associationswere not significant. The pooled OR for each copyof the rs2792 T allele was 1.01 (95% CI, 0.90- 1.13;P 0.89) against cirrhosis controls and 1.04 (95%CI, 0.91- 1.17; P 0.55) against population controls(see Fig. 3).MARC1 (rs2642438)The rs2642438:A variant in MARC1 was consistently less frequent in HCC cases versus controls. Forexample, 24.8% for HCC cases in STOP- HCV versus 29.7% in controls. In regression analysis, the association was relatively weak. The pooled OR for eachcopy of the rs2642438:A allele was 0.91 (95% CI,0.84- 1.00; P 0.043) against cirrhosis controls and*Number of cases indicated here may differ from the number used in regression analyses due to missing data for genotype and/or age, and/or sex.†Control group largely comprises individuals with no history of liver s: hepatitis C- related cirrhosis without HCCSTOP- HCVCases: HCC and hepatitis C- related cirrhosis60.37331.132.214.670128Controls 2: heavy drinkers with neither significant liverdisease nor HCC60.626.432.011.875894Controls 1: alcohol- related cirrhosis without HCC57.125.433.68.99165.01,289Cases: HCC and alcohol- related cirrhosisDresden 5TM6SF2 (rs187429064)1,76,633266INNES ET AL.0.83 (95% CI, 0.72- 0.95; P 0.006) against population controls (see

From the 1School of Health and Life Sciences, Glasgow Caledonian . 23Department of Gastroenterology, Hepatology, and Endocrinology, Hannover Medical School, Hannover, Germany; 24Department of Medicine I, University Hospital Schleswig Holstein . [email protected] Tel.: 44 1